🧠 DocMind

Multi-model PDF assistant — ask anything across your documents.

Live at → multi-docs.streamlit.app

---

What it does

Upload one or more PDFs, pick your LLM provider, and chat with your documents. DocMind chunks and embeds your files locally, builds a vector store, and routes your questions through a RAG chain backed by either Groq or Gemini.

---

Stack

Layer	Tech
UI	Streamlit
Embeddings	HuggingFace (langchain-huggingface)
Vector store	Chroma (langchain-chroma)
LLM providers	Groq · Gemini
PDF loading	PyPDF (pypdf)
Orchestration	LangChain

---

Getting started

1. Clone

git clone https://github.com/your-username/docmind.git
cd docmind

2. Install dependencies

Uses uv for dependency management.

uv sync

3. Run

uv run python -m streamlit run app.py

4. Use

1. Choose a provider — Groq or Gemini
2. Paste your API key
3. Upload your PDFs
4. Click Build pipeline
5. Start asking questions

---

Project structure

docmind/
├── app.py                  # Streamlit UI
├── src/
│   ├── loader.py           # PDF loading
│   ├── splitter.py         # Text chunking
│   ├── embeddings.py       # HuggingFace embeddings
│   ├── vector_store.py     # Chroma vector store + retriever
│   └── rag_chain.py        # RAG chain (Groq / Gemini)
├── pyproject.toml
└── README.md

---

API keys

• Groq → console.groq.com
• Gemini → aistudio.google.com

Ypur API Keys are never stored — entered per session in the UI.

---

Built by Amirtha Ganesh R.